JP5284771B2 - Image compression apparatus and image compression method - Google Patents

Description

  The present invention relates to an image compression apparatus and an image compression method for irreversibly compressing an image input from an image sensor.

  A JPEG (Joint Photographic Experts Group) method is generally known as an image compression method.

  FIG. 5 is a block diagram showing a schematic configuration of a JPEG image compression apparatus. In JPEG, first, a frame image stored in the image data memory 11 is converted into image input data such as 8 × 8 (8 horizontal pixels × 8 vertical lines) by a block forming unit 20A. To read block.

  FIG. 6 is a diagram illustrating an example of image data blocking. As shown in the figure, the first 8 lines (pixel 0 to pixel 7) of the first 8 lines (line 0 to line 7) of the image input data are processed as block 0, and then the block of pixels 8 to 15 is blocked. Processing is performed as 1. In this example, it is assumed that one line is composed of n × 8 pixels (n is an integer). At this time, the start 8 pixels of the next 8 lines (line 8 to line 15) are processed as a block n.

  Next, the blocked data is converted into spatial frequency data by the two-dimensional orthogonal transform unit 30A shown in FIG. As orthogonal transform, JPEG uses two-dimensional DCT (Discrete Cosine Transfer). Further, the spatial frequency data is divided by a desired quantization parameter by the quantization unit 40A to reduce the effective number. Then, the entropy encoding unit 50A performs conversion for assigning data having a high occurrence frequency to a short code length for the quantization result reduced by the quantization unit 40A.

  With the above configuration, lossy compression of image input data is performed. In this case, by changing the above-described quantization parameter and increasing the amount of division, it is possible to further reduce the effective number and realize a high compression rate. However, in JPEG, when the compression rate is increased, block distortion occurs in which the difference in image between blocks increases during expansion. For such a problem, as shown in the prior art encoded data processing apparatus, a method using overlapping orthogonal transform (LOT) in the two-dimensional orthogonal transform unit 30A has been proposed (see, for example, Patent Document 1). ).

  Also in the compression method using overlapping orthogonal transform, the basic configuration of the apparatus is the configuration shown in FIG. 5 as in the case of JPEG. In the overlapped orthogonal transformation method, blocks are compressed so that adjacent blocks have overlapping regions.

  FIG. 7 is a diagram illustrating an example of blocking in the overlapping orthogonal transform. FIG. 7A shows an example in which they overlap in the horizontal direction (direction parallel to the raster scan direction). In this case, in block 0, the data of the first 16 lines of 16 pixels (pixel 0 to pixel 15) are blocked. Next, in the adjacent block 1, the data from the pixel 8 to the pixel 23 is shifted into blocks by 8 pixels. That is, 8 pixels (pixel 8 to pixel 15) overlap. Similarly, a block 2 (not shown) adjacent to the block 1 also overlaps by 8 pixels.

  Next, FIG. 7B shows an example of overlapping in the vertical direction. In this case, in the block n adjacent to the block 0, the data of the line 23 from the line 8 shifted by 8 lines is blocked. Similar to the horizontal direction, eight lines (line 8 to line 15) overlap.

  Next, FIG. 8 shows an example of horizontal and vertical overlapping orthogonal transform processing. FIG. 8A shows an example of horizontal orthogonal orthogonal transform processing. As described above, block 0 and block n are processed in an overlapping manner. An orthogonal transformation process is performed on each of these blocks. At this time, the result of performing horizontal orthogonal transformation on 16 × 16 (16 pixels × 16 lines) is a 16 × 8 block.

  FIG. 8B shows a process of overlapping orthogonal transformation in the vertical direction. Similar to the horizontal direction, the block 0 and the block 1 are processed in an overlapping manner. Orthogonal transformation processing is performed on the 16 × 8 pixel data of each block to obtain 8 × 8 result data. When decompressing, each block becomes an image of 16 × 16 pixels, and an image without block noise can be output by superimposing overlapping portions.

  FIG. 9 shows a block diagram of a two-dimensional orthogonal transform unit 30A using overlapping orthogonal transform disclosed in Patent Document 1 described above. The processing operation in the two-dimensional orthogonal transform unit 30A will be described with reference to the timing chart shown in FIG.

  In this example, the number of lines of the block size (16 lines) of the overlapping orthogonal transform is set as one block line, and the overlapping orthogonal transform is processed in units of block lines.

  First, block line 0 data is input from the preceding block forming unit 20A to the two-dimensional DCT unit 31A, and the two-dimensional DCT unit 31A performs DCT processing on the block line 0 and converts the processing result into a one-dimensional overlapping orthogonal. The data is input to the conversion unit 32A (step S101). The one-dimensional overlapping orthogonal transform unit 32A performs a horizontal compression process (step S102), and writes the result in the buffer memory A (step S103).

  Next, the data of the block line 1 is input to the two-dimensional DCT unit 31A, and the two-dimensional DCT unit 31A performs DCT processing on the block line 1, and inputs the processing result to the one-dimensional overlapping orthogonal transform unit 32A (step). S111). The one-dimensional overlapping orthogonal transform unit 32A performs horizontal compression processing (step S112) and writes the result in the buffer memory B (step S113).

  Next, the one-dimensional overlapping orthogonal transform unit 32A reads the data of the horizontal processing result of the block line 0 stored in the buffer memory A (step S121), and the horizontal processing of the block line 1 stored in the buffer memory B The resulting data is read (step S122), orthogonal transformation processing in the vertical direction is performed on these data (step S123), and the processing result is output to the quantization unit 40A.

  Subsequently, the data of the block line 2 is input to the two-dimensional DCT unit 31A, and the two-dimensional DCT unit 31A performs DCT processing on the block line 2 and inputs the processing result to the one-dimensional overlapping orthogonal transform unit 32A (step) S131). The one-dimensional overlapping orthogonal transform unit 32A performs horizontal compression processing (step S132), and writes the result in the buffer memory A (step S133).

  Next, the one-dimensional overlapping orthogonal transform unit 32A reads the horizontal processing result data of the block line 2 stored in the buffer memory A (step S141), and the horizontal processing of the block line 1 stored in the buffer memory B The resulting data is read (step S142), orthogonal transformation processing in the vertical direction is performed on these data (step S143), and the processing result is output to the quantization unit 40A.

By repeating the above processing operation, the conventional two-dimensional orthogonal transform unit 30A shown in FIG. 8 performs two-dimensional overlapping orthogonal transform on the input image data.
Japanese Patent Laid-Open No. 04-227163

  By the way, in the above-described prior art apparatus of Patent Document 1, when performing the horizontal and vertical overlapped orthogonal transform processing, if the line length is n, the buffer memory 2 surface (16 × n × 2) having a block line size (16 lines). ) Is used. For this reason, there is a problem that the buffer memory size becomes large, and since the buffer memory size is large, there is a problem that power consumption in the image compression apparatus increases.

  In the one-dimensional overlapping orthogonal transform unit, horizontal processing (horizontal compression processing) and vertical processing (vertical compression processing) are alternately performed for each line or for each column. There was a problem that the minute calculation speed decreased.

  The present invention has been made in view of the above problems, and a first object of the present invention is to reduce a buffer memory (a size of a memory that holds an intermediate result of an orthogonal transformation operation) used in an orthogonal transformation operation. An object of the present invention is to provide an image compression apparatus and an image compression method.

  The second object of the present invention is to perform the orthogonal transform operation speed by performing the vertical processing (vertical compression processing) and the horizontal processing (horizontal compression processing) in the overlapping orthogonal transform unit in parallel. An object of the present invention is to provide an image compression apparatus that can be improved.

The present invention has been made to solve the above-described problems, and an image compression apparatus according to the present invention includes:
In order to perform overlapping orthogonal transformation by dividing the raster-scanned image input data into two steps of the vertical direction and the horizontal direction with respect to the raster scanning direction, a predetermined block size for performing the overlapping orthogonal transformation on the image input data is set. a blocking unit for dividing the block unit to be 1/2 of the block size which is the amount of overlap of the horizontal direction, the image input data of the block unit, performs a lapped orthogonal transform operation to said vertical rectangular direction, a first orthogonal transformation unit for storing the operation result in the storage unit, the reading from the first said storage unit a calculation result of the orthogonal transform unit, the second orthogonal transformation unit for performing an overlapping orthogonal transform operation on the horizontal direction If has, the storage unit, the size of the horizontal direction and the vertical direction, the for a given block size, half of the block size respectively the overlap amount And the first orthogonal transform unit outputs a result of performing the orthogonal orthogonal transform operation in the vertical direction in a predetermined order in any of the three divided storage units. The second orthogonal transform unit stores the first orthogonal transform unit among the three divided storage units in parallel with the first orthogonal transform unit performing the overlapping orthogonal transform operation in the vertical direction. The conversion unit reads out the calculation result by the first orthogonal transformation unit from two division storage units different from the division storage unit in which the calculation result is stored, and performs the overlapping orthogonal transformation calculation in the horizontal direction. To do.

The image compression method of the present invention, in order to perform the lapped orthogonal transform in two stages in the vertical direction and the horizontal direction image input data that is raster scanned with respect to the direction of raster scan, lapped orthogonal to said image input data the predetermined block size for conversion, the horizontal direction of the overlap amount is a half of the block procedure as the block size, the the image input data in units of blocks, lapped orthogonal transform operation to said vertical rectangular direction It was carried out, the first orthogonal transformation to save the result to the storing unit, a calculation result by the first orthogonal transformation procedure read from said storage unit, a second performing overlapping orthogonal transform operation on the horizontal direction and orthogonal transformation procedures, but performed by the control unit in the image compression apparatus, wherein the storage unit, the size of the horizontal direction and the vertical direction, with respect to the predetermined block size, respectively It is composed of three divided storage units having a block size of ½, which is the amount of overlap, and the first orthogonal transform procedure is configured to calculate the result of performing the overlap orthogonal transform operation in the vertical direction. The second orthogonal transformation procedure is stored in one of the divided storage units in a predetermined order, and the 3rd orthogonal transformation procedure is performed in parallel with performing the orthogonal orthogonal transformation operation in the vertical direction in the first orthogonal transformation procedure. The calculation results of the first orthogonal transformation procedure are read out from two division storage units different from the division storage unit that stores the calculation results in the first orthogonal transformation procedure among the plurality of division storage units, and the horizontal direction And performing an orthogonal orthogonal transformation operation .

In the image compression apparatus of the present invention, raster-scanned image input data is divided into image data in predetermined block units. The first orthogonal transform unit performs overlapping orthogonal transform operations on the image data in units of blocks in a direction perpendicular to the raster scan direction, and stores the results of the respective operations in the storage unit. The second orthogonal transform unit reads the calculation result in the first orthogonal transform unit from the storage unit, and performs overlapped orthogonal transform computation in the horizontal direction with respect to the raster scan direction.
As a result, the buffer memory used in the orthogonal transform operation (the size of the memory holding the intermediate result of the orthogonal transform operation) can be reduced. For this reason, an image compression apparatus having a smaller configuration can be realized.

  Embodiments of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing a configuration of an image compression apparatus according to an embodiment of the present invention.
In the image compression apparatus 1 shown in FIG. 1, image input data by raster scanning is temporarily stored in the block line memory 10 in order to perform subsequent block unit processing. Other basic configurations are the same as those of the conventional image compression apparatus 1A shown in FIG.

  That is, in the image compression apparatus 1 of the present invention shown in FIG. 1, the blocking unit 20 divides the image data obtained by raster scanning into horizontal and vertical directions in a predetermined unit and blocks the data in predetermined blocks.

  The two-dimensional orthogonal transform unit 30 transforms block unit image data into frequency domain data. Details of the two-dimensional orthogonal transform unit 30 will be described later. The quantization unit 40 performs a process of reducing significant figures from the result of the orthogonal transformation in the two-dimensional orthogonal transformation unit 30 using a predetermined quantization parameter. The entropy encoding unit 50 performs conversion by assigning frequently generated data to a short code length for the quantization result calculated by the quantization unit 40.

  The control unit 60 includes a CPU and the like, and is a control unit that controls a processing operation of each unit in the image compression apparatus 1 to perform a desired processing operation.

  FIG. 2 is a diagram illustrating a configuration example of the two-dimensional orthogonal transform unit 30. As shown in FIG. 2, the two-dimensional orthogonal transform unit 30 includes a vertical overlap orthogonal transform processing unit 31, an intermediate memory unit 32, and a horizontal overlap orthogonal transform processing unit 33.

  In the two-dimensional orthogonal transform unit 30, the blocked image data input from the blocking unit 20 is vertically processed (in a direction perpendicular to the raster scan direction) by the vertical overlap orthogonal transform processing unit 31 in the first stage. The overlap orthogonal transform process is performed, and the result is stored in the intermediate memories 0, 1 and 2 in the intermediate memory unit 32. The horizontal overlap orthogonal transform processing unit 33 then overlaps in the horizontal direction with respect to the data of the vertical overlap orthogonal transform result stored in the intermediate memory unit 32 (data stored in the intermediate memories 0, 1, and 2). Perform orthogonal transform processing. Then, the overlapped orthogonal transformation calculation result is output to the quantization unit 40 at the next stage.

  The image compression apparatus 1 obtains a conversion result of 8 × 8 data by performing overlapping orthogonal transformation on one block (16 pixels × 16 lines) as image input data.

  For this reason, the block forming unit 20 in the image compression apparatus 1 has a block size (8 pixels × 16 lines) of ½ that is the amount of overlap in the horizontal direction with respect to the block size to be subjected to overlapping orthogonal transformation. Block to process. Hereinafter, a half-size block (8 pixels × 16 lines) is referred to as a half block. In FIG. 2, three half blocks 0, 1, and 2 are illustrated.

  FIG. 3 is a timing chart for explaining the processing operation in the two-dimensional orthogonal transform unit 30. The processing operation in the two-dimensional orthogonal transform unit 30 will be described below with reference to the block diagram of FIG. 2 and the timing chart of FIG.

  In the timing chart shown in FIG. 3, the two-dimensional orthogonal transform unit 30 inputs an image of half block 0 (8 pixels × 16 lines) from the preceding block forming unit 20 (step S <b> 11). The vertical overlap orthogonal transform processing unit 31 performs orthogonal transform processing on the image data of the half block 0 (step S12) and stores it in the intermediate memory 0 as 8 × 8 data (step S13).

  Next, the preceding block forming unit 20 inputs the image data of the half block 1 (8 pixels × 16 lines) (step S21). Similarly to the half block 0, the vertical overlap orthogonal transform processing unit 31 performs an orthogonal transform process on the image data of the half block 1 (step S22), and stores it in the intermediate memory 1 as 8 × 8 data (step S23). .

  Further, the preceding block forming unit 20 inputs the image of the half block 2 (step S31). The vertical overlap orthogonal transform processing unit performs orthogonal transform processing on the image data of the half block 2 (8 pixels × 16 lines) (step S32), and stores it in the intermediate memory 2 (step S35) as 8 × 8 data.

  On the other hand, in parallel with the vertical overlap orthogonal transform processing unit 31 performing the processing of the half block 2 (steps S31, S32, S35), the horizontal overlap orthogonal transform processing unit 33 converts the processing result of the half block 0 from the intermediate memory 0. The data is read (step S33), the data of the processing result of half block 1 is read from the intermediate memory 1 (step S34), and the horizontal overlap orthogonal transform processing unit 33 is 1 block (half block 0 + half block 1) data (16 pixels). Orthogonal transformation processing is performed for (× 8 lines) (step S36). As a result, 16 × 8 data is compressed into 8 × 8 data.

  Next, the vertical overlap orthogonal transform processing unit 31 inputs the image data of the half block 3 (step S41), performs orthogonal transform processing on the image data of the half block 3 (step S42), and obtains 8 × 8 data. Store in the intermediate memory 0 (step S43).

  In parallel with the vertical overlap orthogonal transform processing unit 31 performing the orthogonal transform processing (steps S41, S42, and S43) of the half block 3, the horizontal overlap orthogonal transform processing unit 33 converts the half block 1 from the intermediate memory 1 again. Read (step S44), read data of half block 2 from the intermediate memory 2 (step S45), and perform orthogonal transform processing as data (16 pixels × 8 lines) of 1 block (half block 1 + half block 2) (step S46) ). As a result, 16 × 8 data is compressed into 8 × 8 data.

  By repeating the above-described processing, 8 × 8 data conversion results are obtained by performing overlapped orthogonal transformation on one block (16 pixels × 16 lines) of image data for one image.

FIG. 4 is a diagram illustrating a flow of data processing in the two-dimensional orthogonal transform unit 30.
As shown in FIG. 4, the vertical overlap orthogonal transformation processing unit 31 performs vertical processing on data divided in a half block (8 pixels × 16 lines) size in the horizontal direction (horizontal direction in the figure) with the raster direction. An orthogonal transformation calculation process is performed in the direction (step S51), and the data is stored in the intermediate memory as 8 × 8 data (step S52).

  Next, the horizontal overlap orthogonal transform processing unit 33 reads two adjacent data (16 × 8 data) on the intermediate memories 0, 1, 2 and performs orthogonal transform operation processing in the horizontal direction (step S53). . In this case, as shown in the drawing, two pieces of data are read out so that 8 × 8 data are sequentially overlapped, and an orthogonal transformation calculation process is performed in the horizontal direction. In this horizontal overlap orthogonal transform calculation process, the vertical overlap orthogonal transform result of the overlap portion can be reused. For this reason, it is possible to reduce the amount of calculation in the overlapping orthogonal transform process.

  As a result of the horizontal overlap orthogonal transform calculation process, a result of the horizontal overlap orthogonal transform calculation process (8 × 8 data) is obtained (step S54), and the calculation result is output to the quantization unit 40.

  Thus, in the image compression apparatus 1 according to the present embodiment, the data stored in the intermediate memory unit 32 is not in block line units (16 × 8) but in half block units (8 × 8). For this reason, in the case of the present embodiment, the memory capacity (intermediate memory capacity, 8 × 8 × 2 ×) is compared to the memory capacity (buffer memory capacity, 16 × 8 × 2 plane) of the conventional device shown in FIG. 3 planes) can be reduced.

  For the overlapping portion between the blocks in the horizontal direction, the result of the vertical overlapping orthogonal transformation process once calculated is used. That is, when performing horizontal orthogonal orthogonal transformation, the same data is read out from the intermediate memory unit 32 a plurality of times and orthogonal transformation computation is performed, so there is no need to newly compute this portion, and the amount of computation can be reduced. It becomes. Furthermore, since the vertical overlap orthogonal transform process and the horizontal overlap orthogonal transform process can be performed in parallel, the calculation speed can be improved.

  2 shows an example in which three intermediate memories are used. However, only the two intermediate memories 0 and 1 are used as in the conventional two-dimensional orthogonal transform unit shown in FIG. In addition, the memory capacity can be further reduced.

  In this case, first, the vertical duplication orthogonal transformation processing unit 31 performs duplication orthogonal transformation calculation processing in the vertical direction for each of the half blocks 0 and 1, and after storing the processing results in the two intermediate memories 0 and 1 The vertical overlap orthogonal transform processing unit 31 temporarily stops processing. The horizontal overlapping orthogonal transform processing unit 33 performs overlapping orthogonal transform calculation processing in the horizontal direction using the processing results in the vertical direction stored in the two intermediate memories 0 and 1.

  Thereafter, the horizontal overlap orthogonal transform processing unit 33 temporarily stops the process, and the vertical overlap orthogonal transform processing unit 31 performs overlap orthogonal transform calculation processing on the half block 2 in the vertical direction, and stores the processing result in the intermediate memory 0. After saving, the vertical overlap orthogonal transform processing unit 31 temporarily stops the processing. Then, the horizontal overlap orthogonal transform processing unit 33 uses the vertical processing results (processing results of the half block 1 and the half block 2) stored in the two intermediate memories 0 and 1 in the horizontal direction (lateral direction). Performs horizontal overlap orthogonal transformation calculation processing.

  By repeating the above processing procedure, the image data for one image can be compressed. In this case, the memory size of the intermediate memory unit can be greatly reduced as compared with the case of the prior art, but since the number of intermediate memories is two, the vertical overlap orthogonal transform processing unit 31 and the horizontal overlap orthogonal transform processing unit 33, a pause period occurs, and the calculation processing speed is reduced as compared with the case where three intermediate memories are used.

  As mentioned above, although embodiment of this invention was described, various deformation | transformation and a change are naturally possible for each structure of embodiment mentioned above. For example, the block size and overlap size are not limited to 16 pixels and 8 pixels.

  In the above-described embodiment, the first orthogonal transform unit described above corresponds to the vertical overlap orthogonal transform processing unit 31 illustrated in FIG. The storage unit described above corresponds to the intermediate memory unit (intermediate memory 0, 1, 2) 32. The second orthogonal transform unit described above corresponds to the horizontal overlap orthogonal transform processing unit 33.

  The image compression apparatus 1 according to the present embodiment includes a blocking unit 20 that divides raster-scanned image data into image data in predetermined block units, and raster-scanned image data in the raster scan direction. On the other hand, a vertical orthogonal orthogonal transform processing unit 31 that performs an overlapping orthogonal transform operation in a predetermined block unit in a vertical direction and stores the operation result in the intermediate memory unit 32, and an operation result by the vertical overlapping orthogonal transform processing unit 31 in the intermediate memory A horizontal overlapping orthogonal transformation processing unit 33 that reads out from the unit 32 and performs overlapping orthogonal transformation calculation in the horizontal direction with respect to the raster scan direction.

  Thereby, compared with the conventional apparatus shown in FIG. 9, orthogonal transformation processing can be performed with a smaller buffer memory. For this reason, the scale of the circuit of the image compression apparatus can be reduced, and power consumption in the image compression apparatus can be reduced. In addition, since the vertical overlap orthogonal transform processing unit 31 and the horizontal overlap orthogonal transform processing unit 33 can perform processing in parallel at the same time, the calculation speed can be improved.

  Although the embodiments of the present invention have been described above, the above-described image compression apparatus 1 has a computer system therein. Each processing unit constituting the image compression apparatus 1 may be realized by dedicated hardware, and a program for realizing the function of each processing unit is recorded on a computer-readable recording medium. Then, the program recorded on the recording medium may be read into a computer system and executed to realize the function. Here, the “computer system” includes an OS and hardware such as peripheral devices.

  The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system.

  Although the embodiment of the present invention has been described above, the image compression apparatus of the present invention is not limited to the above illustrated example, and various modifications can be made without departing from the scope of the present invention. Of course.

It is a figure which shows the structure of the image compression apparatus concerning embodiment of this invention. It is a figure which shows the structural example of a two-dimensional orthogonal transformation part. It is a timing chart for demonstrating the processing operation in a two-dimensional orthogonal transformation part. It is a figure which shows the flow of the data processing in a two-dimensional orthogonal transformation part. It is a figure which shows schematic structure of the image compression apparatus of a JPEG system. It is a figure which shows the example of blocking of image data. It is a figure which shows the example of blocking in overlap orthogonal transformation. It is a figure for demonstrating the horizontal and orthogonal overlap orthogonal transformation process. It is a figure which shows the structure of the two-dimensional orthogonal transformation part using the overlapping orthogonal transformation of a prior art. 10 is a timing chart showing processing operations in the two-dimensional orthogonal transform unit shown in FIG. 9.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1,1A ... Image compression apparatus, 10 ... Block line memory, 11 ... Image data memory, 20, 20A ... Blocking part, 30, 30A ... Two-dimensional orthogonal transformation part, 31. .. Vertical overlapping orthogonal transformation processing unit, 31A ... Two-dimensional DCT unit, 32 ... Intermediate memory unit, 32A ... One-dimensional overlapping orthogonal transformation unit, 33 ... Horizontal overlapping orthogonal transformation processing unit, 40, 40A: quantization unit, 50, 50A: entropy coding unit

Claims (2)

In order to perform overlapping orthogonal transformation by dividing the raster-scanned image input data into two steps of the vertical direction and the horizontal direction with respect to the raster scanning direction, a predetermined block size for performing the overlapping orthogonal transformation on the image input data is set. a blocking unit for dividing the block unit to be 1/2 of the block size the a redundant amount of horizontal,
The image input data of the block unit, performs a lapped orthogonal transform operation to said vertical rectangular direction, a first orthogonal transformation unit for storing the operation result in the storage unit,
Read the operation result by the first orthogonal transformation unit from the storage unit, a second orthogonal transformation unit for performing an overlapping orthogonal transform operation on the horizontal Direction,
Have
The storage unit is composed of three divided storage units in which the horizontal and vertical sizes are each ½ the block size that is the overlap amount with respect to the predetermined block size,
The first orthogonal transform unit stores a computation result obtained by performing an overlapped orthogonal transform computation in the vertical direction in one of the three divided storage units in a predetermined order;
The second orthogonal transform unit includes the first orthogonal transform unit that includes the three orthogonal storage units in parallel with the first orthogonal transform unit performing the overlapping orthogonal transform operation in the vertical direction. Reading out the calculation result by the first orthogonal transform unit from two divided storage units different from the divided storage unit for storing the calculation result, and performing an overlapping orthogonal transform operation in the horizontal direction;
An image compression apparatus characterized by the above. In order to perform overlapped orthogonal transformation on raster scanned image input data in two stages in a vertical direction and a horizontal direction with respect to the direction of raster scanning, a predetermined block size for performing overlapped orthogonal transformation on the image input data , and block procedure which is 1/2 of the block size which is the amount of overlap in the horizontal direction,
The image input data of the block unit, performs a lapped orthogonal transform operation to said vertical rectangular direction, a first orthogonal transformation To save the result to the storage unit,
Read the operation result by the first orthogonal transformation procedure from the storage unit, and the second orthogonal transformation procedure for lapped orthogonal transform operation to said horizontal Direction,
Is performed by the control unit in the image compression apparatus ,
The storage unit is composed of three divided storage units in which the horizontal and vertical sizes are each ½ the block size that is the overlap amount with respect to the predetermined block size,
In the first orthogonal transform procedure, the operation result obtained by performing the overlapping orthogonal transform operation in the vertical direction is stored in a predetermined order in any of the three divided storage units,
The second orthogonal transformation procedure is the first orthogonal transformation procedure of the three divided storage units in parallel with performing the orthogonal orthogonal transformation operation in the vertical direction in the first orthogonal transformation procedure. Reading out the calculation result by the first orthogonal transformation procedure from two division storage units different from the division storage unit for storing the calculation result, and performing the overlapping orthogonal transformation calculation in the horizontal direction,
An image compression method characterized by the above.

Images